Stream-based parallel computation for discrete wavelet transform by using graphics hardware

The discrete wavelet transform (DWT) has been widely used in various scientific and engineering fields. However, the enormous computation of DWT caused by multilevel filtering/down-sampling is a bottleneck that limits the application of DWT used in real-time environment where the data size is large. A stream-based parallel computation framework to accelerate the implementation of DWT is presented in this paper, which is based on employing the consumer-level programmable graphics hardware. Simulation results show that, this stream-based parallel computation framework can achieve a significant performance gain on algorithm acceleration comparing with those completely CPU-based solutions for DWT.